Bioinformatics Seminars

Bioinformatics Seminar

Time: 11AM
Venue: Zoom Webinar

14 September 2021

Robust compositional analysis of single-cell data

Stefano Mangiola
WEHI Bioinformatics

Single-cell transcriptomics allows the unbiased characterisation of the cellular composition of tissues. The cellular composition can be compared between biological or clinical conditions to identify potential cellular drivers. This strategy has been critical to unveil drivers of immune response in cancer and pathogen infection from single-cell data. Developing a robust statistical method for differential composition analyses from single-cell data is crucial for driving discoveries. The compositional data from single-cell experiments has four main properties. The data is in count form; counts underlie inversely correlated proportions that sum to one; larger cell groups are more variable across samples than small groups; real-world data is rich in outlier observation. A model that covers all these properties is currently lacking. Here, I present a robust and outlier-aware method for testing differential tissue composition from single-cell data. This model can also transfer knowledge from a large set of integrated datasets to increase accuracy further. I present how this model can be applied to identify drivers in metastatic breast cancer.

Search past seminars